Upload 141-Electrify.mp4
Problem Statement
High-voltage electrical work is dangerous.
Training robots to handle real-world circuitry could reduce human risk — but the data to do so doesn't exist.
Nearly all datasets involving electrical manipulation are closed, proprietary, or simulated.
There's no open-access dataset available for training robotic systems to physically interact with live circuits. The confidentiality of datasets limits growth in open-source robotic intelligence.
Solution:
We present an open-source dataset to train the SO-101 ARM for circuit connection. Our dataset allows the SO-101 to open alligator clips, connect them to circuit terminals, and pick up and place batteries into the power source to build a functional circuit and power light bulbs.
We worked on the Imitation Learning Track 1.
Datasets:
Alligator Clips:
LeRobot-worldwide-hackathon/141-Electrify-open_alligator
Battery Placement:
LeRobot-worldwide-hackathon/141-Electrify-battery_placing
LeRobot-worldwide-hackathon/141-Electrify-insertbattery
Models
Alligator Clips: https://huggingface.co/Ganga008/allegator_place_orange
Battery Placement: https://huggingface.co/ronantakizawa/battery-placement